|dc.description.abstract||This dissertation comprises three essays, relating to negative externalities in economics. The first essay concentrates on residential electricity consumption. In the economic literature, price elasticity of demands estimates for this market vary widely. In this essay, I seek to explain these findings using three nationwide datasets – the American Housing Survey, Forms EIA-861, and the Residential Energy Consumption Survey – from the U.S. I examine the role of the sample period, level of aggregation, use of panel data, use of instrumental variables, and inclusion of housing characteristics and capital stock. The findings suggest that price elasticities have remained relatively constant over the time period considered, from 1997 to 2009. Upon splitting my panel datasets into annual cross sections, I do observe a negative relationship between price elasticities and the average price. Whether prices are rising or falling appears to have little effect on the estimates. I also find that aggregating our data can result in both higher and lower price elasticity estimates, depending on the dataset used, and that controlling for unit-level fixed effects with panel data generally results in more inelastic demand functions. Addressing the endogeneity of price and/or measurement error in price with instrumental variables has a small but noticeable effect on the price elasticities. Finally, controlling for housing characteristics and capital stock produces a lower price elasticity.
My second essay focuses on personal vehicular transportation, which is the source of several externalities, including congestion and conventional air pollutant and greenhouse gas emissions. In this work, I examine the geographical distribution of carbon dioxide emissions from cars. Specifically, I focus on rural and urban households in the UK, using repeated cross sections from the UK National Travel Survey. I contrast driving behavior with new vehicle purchases, and ask three related questions: First, do rural households purchase more vehicles, and/or vehicles with higher emissions rates? Second, do they drive more miles than their urban counterparts? Third, how do the “carbon footprints” from these two groups compare? To answer these questions, I first model the number of vehicles chosen by each household, and the emissions type of each. I then model the travel demand for that vehicle, conditioning on the latter choice. I use these results to estimate the contribution of rural and urban residents to total CO2 emissions. I find that rural households own more vehicles than urban households, and that these vehicles have higher emissions rates. Rural vehicles are also driven 12.9% more than urban vehicles. Lastly, I estimate a carbon footprint that is 58% higher for rural households than urban.
My third essay considers an application of the hedonic price models, which are widely used in nonmarket valuation to find the value of environmental quality, changes in health risks, etc. The approach relies on property values. It is less commonly employed, however, to assess the value of certain amenities like crime mitigation. One problem with deploying hedonics in this area is that crime tends to be correlated with unobserved neighborhood attributes, and for that reason regression coefficients on crime may be biased. This chapter considers one particular set of crimes, methamphetamine laboratories, on property values. While methamphetamine labs are more likely to be established and discovered in some neighborhoods more than others, I follow previous literature by arguing that, within the confines of the neighborhoods they are discovered in, they are as good as randomly assigned and can thus be regarded as locally exogenous in regression analysis. I use three registers – one federal, two local – to identify the locations of past discoveries, in Summit County, OH, and categorize the discoveries by scale (e.g., a small discovery of meth in a dumpster versus an actual meth lab). I find mixed evidence that the value of homes closest to a discovery are negatively affected, either directly or in terms of relatively higher properties for properties that are somewhat further away from such discoveries. The effect, when identified, is slightly stronger for large-scale discoveries. I also consider the effect of information disclosure, and find that, for repeat sales observations, property sales that occur following a discovery within 200 meters and after public registers are available observe a sizable loss in value. The robustness each of these findings, however, appears questionable, since their magnitude and statistical significance are sensitive to model specification.||en_US